Re: [PATCH 2/2] Handle group offsets in UFS1

[Robert Millan]

On Thu, Aug 13, 2009 at 10:25:15PM +0200, Vladimir 'phcoder' Serbinenko wrote:
> On Thu, Aug 13, 2009 at 10:05 PM, Robert Millan<address@hidden> wrote:
> > On Thu, Aug 13, 2009 at 06:04:35PM +0200, Vladimir 'phcoder' Serbinenko 
> > wrote:
> >
> > This looks a bit odd (a mask applied to an integer?), but if it's really
> > this way, please go ahead with it.
> >
> It's the so-called skewness.
> Let's say you place inodes on addresses (C/H/S)
> (0/0/1) and (1/0/1) you first read the metadata at (0/0/1) then you
> try to fetch the metadata from (0/1/1). Responding to your request
> harddrive moves the head to cylinder number 1 but it takes some time.
> Meanwhile the plates have spinned (they are spinning constantly) and
> perhaps head is above sector (1/0/10) and you need to wait for
> complete rotation to fetch your sector.
> If you write inodes at (0/0/1) and (1/0/15) you will need to wait only
> for 4 sectors. This is called skewness and was an optimisation
> technique in the past. But now OS doesn't know about physical geometry
> and so can't do such kind of optimisation. I suppose it's why it's not
> used anymore for UFS2. I don't know if FreeBSD variant of UFS1 still
> uses this feature.

I see :-)

-- 
Robert Millan

  The DRM opt-in fallacy: "Your data belongs to us. We will decide when (and
  how) you may access your data; but nobody's threatening your freedom: we
  still allow you to remove your data and not access it at all."